impala competence center asset management - cost driven risk management

16
White Paper Impala Competence Center Asset Management Investment Analysis – Life Cycle Cost Analysis: Cost Driven Risk Management Benny Lauwers Consultancy Manager Impala

Upload: impala-consulting

Post on 25-Dec-2014

217 views

Category:

Economy & Finance


3 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Impala competence center   asset management - cost driven risk management

White Paper Impala Competence Center Asset Management

Investment Analysis – Life Cycle Cost Analysis:

Cost Driven Risk Management

Benny Lauwers Consultancy Manager Impala

Page 2: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 2 of 16

Content    

1.   Introduction ............................................................................................................. 3  

2.   Risk matrices........................................................................................................... 3  

3.   Converting the risks in a cost model .................................................................... 4  

4.   The notion “risk cost” ............................................................................................ 7  4.1.   Costs that are charged ....................................................................................... 7  4.2.   Determination of the year where the costs are charged..................................... 8  

5.   Decision ................................................................................................................. 15        

Page 3: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 3 of 16

 1. Introduction

Risk  management  involves  not  only  the  evasion  of  risks,  but  also  the  acceptance  of  certain  risks.    

The  trend  of  accepting  and  controlling  risks  instead  of  merely  reducing  them  has  been  going  for  some  

time.    

In  the  world  of  asset  management  risk  analyses  are  often  used  to  determine  the  critical  failure  modes  of  an  asset,  and  how  these  failure  modes  can  be  made  less  critical.  In  this  case  critical  is  defined  as  a  

situation  where  the  failure  of  an  asset  and  the  consequences  for  the  company  are  viewed  as  hazardous.  

Failure  modes  can  be  made  less  critical  by  reducing  the  frequency  and  consequences  of  the  failure.  This  can  be  done  by  changing  the  asset  and  making  use  of  more  reliable  components.  Another  method  is  

setting  up  a  predictive  maintenance  programme  that  shows  the  erosion  or  wear,  and  can  enable  action  before  the  asset  fails.  Throughout  the  white  paper  the  word  ‘action’  will  be  used.  Please  keep  in  mind  that  this  concept  can  be  interpreted  in  different  manners  in  practise.    

The  methodology  that  will  be  described  in  this  paper  can  provide  support  in  the  decision  whether  an  

action  should  be  taken  based  on  life  cycle  cost.  The  life  cycle  cost  is  based  on  the  determination  of  the  cost  of  risk,  which  is  the  product  of  the  probability  of  failure  and  the  costs  of  the  failure  consequences.    

2. Risk matrices

When  executing  a  risk  analysis,  often  use  is  made  of  risk  matrices.  A  risk  matrix  shows  the  probability  of  

failure  with  reference  to  the  consequences  of  the  failure.  The  combination  fo  this  probability  and  consequence  provides  an  analysed  failuremode  in  a  specific  place  in  the  risk  matrix.  The  risk  matrix  often  shows  a  critical  zone,  where  the  frequency  of  the  failure  and/or  the  consequences  are  very  large;  

and  a  non-­‐critical  zone  where  they  are  small.  The  critical  and  non-­‐critical  zones  are  seperated  by  a  criticalityborder.    

In  some  risk  matrices  one  can  find  several  degrees  or  levels  in  the  transition  of  the  critical  to  non-­‐critical  

zone.  In  this  paper  we  will  confine  ourselves  to  a  simple  matrix,  with  a  critical  and  non-­‐critical  zone,  but  please  be  aware  that  the  reasoning  is  identical  to  the  risk  matrices  that  have  several  degrees  or  levels.    

  Consequence:  Unavailability  /  Failure  of  the  asset  Probability  of  failure  

(1/x)  16  hours   8  hours   4  hours   1  hour  

6  months  (1/2)         Non-­‐critical  zone  

3  months  (1/4)          1  month  (1/12)          1  week  (1/52)   Critical  zone        

Figure 1: Example of a risk matrix with a critical zone and a non-critical zone

The  classical  method  of  working  when  a  failuremode  enters  the  critical  zone  is  that  several  actions  will  

have  to  be  undertaken  to  ensure  that  the  failuremode  is  moved  from  this  zone.  This  can  be  done  by  

Page 4: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 4 of 16

taking  preventive  actions,  that  lower  the  probability  of  failure,  but  it  is  also  possible  to  take  damage  reducing  actions,  that  lower  the  consequences  of  the  failure.  The  preference  is  usually  given  to  the  

preventie  actions,  before  taking  damage  reducing  measures.    

Currently,  however,  does  the  undertaking  of  these  actions  involve  an  evaluation  of  the  costs  of  these  actions?  And  to  what  extent  are  they  compared  to  the  added  value  that  the  actions  will  bring?  Usually  

neither  of  these  questions  are  asked.    

In  many  cases  a  convulsive  search  for  solutions  is  done  that  will  keep  the  failuremode  outside  of  the  critical  zone,  regardless  the  costs  of  these  solutions.  Experience  teaches  us  that  it  can  be  more  cost  

efficitent  to  accept  the  failuremode  in  the  critical  zone,  as  the  costs  of  the  actions  do  not  outweigh  their  actual  financial  added  value.  This  is  known  as  ‘cost  driven  risk  management’.    

The  method  that  uses  life  cycle  costs  to  decide  whether  an  action  can  be  taken  will  be  described  

further,  and  is  the  subject  matter  of  this  paper.  This  methodology  can  also  be  used  to  compare  several  different  possible  actions,  and  to  enable  the  choice  for  an  action  with  the  highest  added  value.    

3. Converting the risks in a cost model

As  was  described  in  the  introduction,  a  risk  matrix  is  a  summary  of  the  possible  probabilities  and  the  probable  consequences.    

An  assumption  is  made  that  every  consequence,  such  as  a  production  stagnation  of  a  specific  time,  a  

quality  problem,  an  environmental  or  safety  incident,  etc.  can  be  converted  to  a  specific  cost.    

When  we  set  up  a  matrix,  where  the  consequence  is  shown  as  a  function  of  the  cost,  and  we  combine  

the  probability  of  failure  with  these  consequence  costs,  it  is  known  as  the  ‘risk  costs’.    

The  matrix  can  then  be  set  up  as  follows:  

RC  =  Risk  costs=  Probability  x  Costs  

(in  this  example  we  use  costs  on  a  yearly  basis  to  simplify  the  matrix  \)  

  Consequence:  Unavailability  /  Failure  of  the  asset  Probability  of  failure  

(1/x)  16  hrs  =  3.200  €   8  hrs  =  1.600  €   4  hrs  =  800  €   1  hrs  =  200  €  

6  months  (1/2)   RK  =  2  x  3.200€   RK  =  2  x  1.600€   RK  =  2  x  800€   RK  =  2  x  200€  3  months  (1/4)   RK  =  4  x  3.200€   RK  =  4  x  1.600€   RK  =  4  x  800€   RK  =  4  x  200€  1  month  (1/12)   RK  =  12  x  3.200€   RK  =  12  x  1.600€   RK  =  12  x  800€   RK  =  12  x  200€  1  week  (1/52)   RK  =  52  x  3.200€   RK  =  52  x  1.600€   RK  =  52  x  800€   RK  =  52  x  200€  

Figure 2: Example of a risk matrix where probability is set out against consequencecost

The  risk  costs  in  the  red  area  will  not  be  accepted,  and  is  defined  as  critical  from  here  on.  The  risk  costs  

in  the  green  area  are  seen  as  acceptable  and  are  not  seen  as  critical    

Page 5: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 5 of 16

According  to  the  commonly  used  methodologies  we  would  like  to  emphasise  that  when  analysis  shows  that  the  critical  zone  has  been  reached,  an  action  will  be  necessary  to  ensure  that  the  critical  zone  will  

be  exited.    

Through  taking  actions  we  will  attempt  to  lower  the  risk  costs  in  such  a  manner  that  it  becomes  acceptable.  An  important  aspect  to  mark  is  that  any  action  that  will  be  taken  will  also  involve  action  

costs.    

  Consequence:  Unavailability  /  Failure  of  the  asset  Probability  of  failure  

(1/x)  16  hrs  =  3.200  €   8  hrs  =  1.600  €   4  hrs  =  800  €   1  hrs  =  200  €  

6  months  (1/2)     RC  =  2  x  1.600€      3  months  (1/4)       Cost  of  action    1  month  (1/12)          1  week  (1/52)     RC  =  52  x  1.600€      

Figure 3: Lowering the riskcosts + costs of action

A  cost  efficient  action  thus  is  an  action  where  the  costs,  added  to  the  new  acceptable  risk  costs,  are  

lower  than  the  original,  unacceptable  risk  costs.    

This  can  also  be  expressed  as  follows:  

Cost  of  action  +  Non-­‐critical  risk  cost  <  Critical  risk  cost  

Or  

CA  +  CNC  <  CC   (1)  

Ca  =  Cost  of  the  action  CNC  =  Non-­‐critical  risk  cost  CC  =  Critical  risk  cost  

When  we  look  at  the  example  (figure  3)  this  becomes:  

Cost  of  the  Action  +  RC:  2  x  1.600€  <  RC  52  x  1.600€  

If  the  critical  risk  costs  turn  out  to  be  lower  than  the  non-­‐critical  risk  costs  added  to  the  cost  of  action,  one  can  question  the  added  value  of  the  action.  It  may  be  better  to  (temporarily)  accept  the  risk  costs.  It  is  evident  that  a  strategic  choice  can  be  made  to  undertake  the  action,  regardless  of  the  cost  

efficiency  of  the  action.    

In  the  reasoning  so  far  we  have  assumed  a  risk  cost  that  can  be  lowered  on  one  occasion  by  taking  a  specific  action.    

In  structural  asset  management,  however,  it  is  important  that  the  costs  of  an  action  are  specified  for  

the  total  life  cycle  of  the  asset.  The  action  costs  will  have  to  be  written  off  over  the  remaining  life  time  

that  the  asset  will  be  in  use.    

Page 6: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 6 of 16

This  is  why  the  notion  ‘Net  Present  Value’  will  be  introduced.  The  Net  Present  Value  (NPV)  takes  into  account  all  the  different  cash  flows,  coupled  to  the  asset,  that  flow  in  and  out  in  a  number  of  years.  The  

NPV  discounts  these  future  cash  flows  to  their  current  or  present  value.  The  sum  of  all  these  discounted  yearly  cash  flows  is  the  Net  Present  Value.    

  (2)  

Where:  

Ct  =  the  cash  flow  in  year  t  t  =  the  year  

r  =  discount  rate  /  weighed  average  capital  cost    

It  is  necessary  to  discount  to  present  value  to  ensure  that  inflation  is  taken  into  account,  but  also  to  realise  a  minimum  return  on  investment,  in  this  case  for  an  action.  This  is  the  so-­‐called  Weighted  

Average  Cost  of  Capital  (WACC).  Usually  this  can  be  requested  at  the  financial  department  of  an  organisation.    

To  discount  the  value  of  a  future  cash  flow  in  year  t  to  present  value,  the  following  formula  can  be  used.    

  (3)  

PV  =  Present  Value;  Current  value  of  future  cash  flow  

To  determine  the  NPV  both  incoming  and  outgoing  cash  flows  are  necessary.  In  this  analysis  we  assume  that,  regardless  of  the  action,  the  incoming  cash  flows  will  remain  equal.  In  practise,  these  cash  flows  

will  increase  due  to  the  fact  that  an  asset  will  attain  a  higher  availability.  For  this  analysis  this  is  not  relevant.    

In  this  analysis  we  will  thus  work  solely  with  outgoing  cash  flows.  Moreover,  we  will  only  work  with  

specific  consequence  costs  (deprivation  costs)  and  the  costs  of  action.  We  assume  that  the  other  outgoing  cash  flows  will  remain  the  same,  despite  the  action.    

Please  be  aware  that  we  will  be  using  positive  values  in  our  calculations,  where  outgoing  cash  flows  and  

thus  negative  values  should  be  used.    

By  using  the  Net  Present  Value,  the  formula  for  critical  action  (1)  can  be  rewritten:  

NPVA  +  NPVNC  <NPVC   (4)  

NPVa  =  Cost  of  the  action  

NPCNC  =  Non  critical  risk  cost  NPVC  =  Critical  risk  cost  

Page 7: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 7 of 16

In  this  calculation  the  Net  Present  Value  of  the  action  is  also  taken  into  account,  as  actions  do  not  always  have  to  take  place  straight  away  and  can  be  planned  in  future  years.  The  action  can  be  taken  in  a  

specific  year,  or  can  be  spread  over  several  years,  which  means  that  depreciation  has  to  be  taken  into  account.    

On  the  basis  of  the  total  NPV,  the  left  side  of  the  formula  (4)  can  show  the  optimal  timing  for  the  

implementation  of  a  single  action.  This  is  the  year  that  shows  the  lowest  NPV.  Several  scenarios  thus  need  to  be  calculated,  where  the  year  the  action  is  taken  varies.    

Example:  If  the  remaining  life  cycle  of  an  asset  is  another  20  years,  20  scenarios’  can  be  calculated,  

where  the  cost  of  the  action  is  billed  in  different  years.  The  scenario  with  the  lowest  total  NPV  is  the  year  that  the  action  should  be  implemented,  according  to  cost  strategy.  In  the  period  of  time  before  the  implementation  of  the  action,  the  ‘old’  risk  cost  has  to  be  taken  into  account;  while  the  ‘new’  risk  cost  

has  to  be  taken  into  account  after  the  action  is  completed.    

 4. The notion “risk cost”

As  has  been  mentioned  before,  risk  cost  is  the  product  of  consequence  cost  and  probability.  Both  

parameters  are  elements  of  the  risk  matrix.  The  probability  can  often  be  found  immediately,  and  the  consequences  can  then  be  adjusted  to  costs.    

The  probability  that  something  will  happen  is  a  function  of  the  failure  pattern  of  the  asset.  The  determination  of  this  failure  pattern  will  be  described  extensively  in  this  document.    

At  this  point,  the  determination  of  the  specific  costs  that  are  used  in  the  calculations  are  key,  as  well  as  

the  year  in  which  these  costs  are  effectively  booked  or  charged.    

4.1. Costs that are charged

The  following  costs  are  components  of  the  total  factor  costs:  

Direct  costs  =  maintenance  costs:  

• Direct  labour  costs  as  a  result  of  yearly  maintenance  activities  • Direct  material  costs  as  a  result  of  yearly  maintenance  activities  

There  is  a  chance  that  these  costs  will  be  different  before  and  after  the  action  has  been  taken.  An  

example  would  be  when  a  maintenance  plan  is  altered  of  made  more  specific,  which  could  lead  to  the  addition  or  removal  of  maintenance  activities.    

Indirect  cost  =  deprivation  cost:  

• The  consequence  costs  that  are  coupled  to  the  failure  of  the  asset.  This  could  be  losses  due  to  loss  of  production,  costs  coupled  to  an  accident  etc.    

Here  it  is  also  possible  that  these  costs  differ  before  and  after  taking  the  action.    

Page 8: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 8 of 16

4.2. Determination of the year where the costs are charged

The  direct  costs  for  a  specific  year  can  be  determined  and  assigned  very  easily.  This  a  summation  of  the  maintenance  costs,  and  possible  labour-­‐  and  material  costs  of  longer  shutdowns  and  revisions.    

In  the  example  shown  below,  the  yearly  maintenance  costs  remain  the  same.  The  exception  is  a  3-­‐

yearly  revision  which  leads  to  higher  costs  in  those  specific  years  (3,  6,  9,  12  …).    

Total  maintenance  costs  per  year  

Year  1   Year  2   Year  3   Year  4   Year  5  

Direct  labour   6.200  €   6.200  €   9.400  €   6.200  €   6.200  €  Direct  material   1.300  €   1.300  €   2.900  €   1.300  €   1.300  €  Total  maintenance  cost   7.500  €   7.500  €   12.300  €   7.500  €   7.500  €  PV  maintenance  cost  @  WACC  7%  

7.009,35  €   6.550,8  €   10.040,5  €   5.721,75  €   5.347,4  €  

Table 1: Total yearly maintenance costs (PV: Present Value; WACC: Weighted Average Cost of Capital)

Remark:  The  current  costs  (year  0)  are  not  taken  into  account  in  this  calculation.  The  option  exists  to  incorporate  these  costs.  When  they  are  incorporated  they  should  not  be  discounted  as  they  are  

(logically)  at  present  value.  

The  difficulty  in  this  case  is  the  determination  of  the  year  in  which  the  deprivation  costs  should  be  charged,  as  this  is  a  function  of  the  failure  pattern  of  the  asset.    

There  are  two  ways  in  which  this  can  be  determined:  

• A  deterministic  method:  Simple,  but  not  very  accurate    • A  stochastic  method:  A  little  more  complex,  but  more  accurate    

In  both  cases  the  starting  point  is  the  Mean  Time  To  Failure  (MTTF)  or  Mean  Time  Between  Failure  

(MTBF).  The  theoretical  discussion  of  the  difference  between  the  two  is  not  shown  here.    

4.2.1. Determination  of  the  failure  pattern  according  to  the  deterministic  method    

In  this  case  the  deterministic  method  assumes  that  we  know  when  an  asset  will  fail.  In  practise  this  is  not  realistic,  as  we  will  never  know  when  an  asset  is  going  to  fail.  This  is  impossible  to  predict.    

We  assume  that  the  value  of  the  MTTF  is  the  moment  that  the  asset  will  actually  fail,  even  though  this  

is  an  average  value  in  reality.  This  is  the  reason  that  the  deterministic  method  is  the  least  accurate  method.    

This  method  is  very  suitable  to  give  a  simple  and  fast  indication.    

The  MTTF  can  help  determine  the  number  of  failures  per  year.  The  table  shows  these  values  for  different  values  of  the  MTTF.    

Page 9: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 9 of 16

 Number  of  failures  per  year  

Year  1   Year  2   Year  3   Year  4   Year  5   Year  6   Year  7  

MTTF  =  1  year   1   1   1   1   1   1   1  MTTF  =  3  year   0   0   1   0   0   1   0  MTTF  =  0,5  year   2   2   2   2   2   2   2  MTTF  =  1,2  year   0   1   1   1   1   1   0  

Table 2: Examples of a failure pattern in the function of the MTTF

A  deprivation  cost  can  now  be  assigned  to  every  failure.  The  table  shown  below  shows  the  complete  

cost  overview  per  year,  taking  into  account  the  direct  and  indirect  costs.  In  this  example  we  use  a  deprivation  cost  of  €  25.000  and  an  MTTF  of  1,2  years.    

Total  costs  per  year   Year  1   Year  2   Year  3   Year  4   Year  5  Direct  labour   6.200  €   6.200  €   9.400  €   6.200  €   6.200  €  Direct  material   1.300  €   1.300  €   2.900  €   1.300  €   1.300  €  Total  maintenance  costs   7.500  €   7.500  €   12.300  €   7.500  €   7.500  €  Deprivation  costs  @  MTTF  =  1,2  

0  €   25.000  €   25.000  €   25.000  €   25.000  €  

Total  cost   7.500  €   32.500  €   37.300  €   32.500  €   32.500  €  PV  cost  @  WACC  7%   7.009,35  €   28.386,76  €   30.447,91  €   24.794,09  €   23.172,05  €  NPV   113.810,16  €          

Table 3: Total cost per year with a deterministic failure pattern and an MTTF of 1,2 years

Pay  attention:  This  calculated  Net  Present  Value  shows  only  the  outgoing  cash  flow,  more  specifically  

the  direct  costs  related  to  maintenance  and  the  deprivation  costs  related  to  the  failure  of  the  asset.    

If  the  MTTF,  the  costs  of  the  action  and  the  direct  and  indirect  costs  are  known  for  the  situation  before  taking  the  action  and  after  taking  the  action,  one  can  compare  the  NPV’s  of  both  situations.    

On  the  basis  of  these  calculated  NPV’s  one  can  decide  whether  the  action  should  be  implemented.  In  

the  example  shown  in  the  table  above  an  MTTF  of  1,2  years  was  used.  If  we  label  that  example  as  the  current  situation,  and  assume  that  the  action  ensures  that  the  MTTF  has  increased  to  4  years.  We  

assume  that  the  direct  cost  and  the  deprivation  costs  remain  unchanged  after  taking  the  action.  The  situation  after  taking  the  action  is  shown  in  the  next  table.    

Total  costs  per  year   Year  1   Year  2   Year  3   Year  4   Year  5  Direct  labour   6.200  €   6.200  €   9.400  €   6.200  €   6.200  €  Direct  material   1.300  €   1.300  €   2.900  €   1.300  €   1.300  €  Total  maintenance  costs   7.500  €   7.500  €   12.300  €   7.500  €   7.500  €  Deprivation  costs  @  MTTF  =  1,2  

0  €   0  €   0  €   25.000  €   0  €  

Total  cost   7.500  €   7.500  €   12.300  €   32.500  €   7.500  €  PV  cost  @  WACC  7%   7.009,35  €   6.550,8  €   10.040,5  €   24.794,09  €   5.347,4  €  NPV   53.742,14  €          

Table 4: Total cost per yearwith a deterministic failure pattern and an MTTF of 4 years

Page 10: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 10 of 16

Comparing  the  NPV’s  provides  the  following  results:  

NPV  before  taking  the  action:   113.810,16  €  NPV  after  taking  the  action:   53.742,14  €  Result:   IMPLEMENT  ACTION  

Table 5:Comparison of NPV’s before and after taking the action

When  we  want  to  execute  a  single  action  in  this  example,  such  as  a  modification,  with  a  cost  of  €  

12.500,  we  can  use  the  data  to  assess  several  scenarios.  In  the  scenario’s  the  action  will  be  taken  in  different  years.  The  scenario  with  the  lowest  total  NPV  shows  the  year  that  the  action  can  best  be  implemented.    

NPV  scenario  year  1:     65.424,33  €   NPV  scenario  year  2:     86.496,04  €  NPV  scenario  year  3:     116.229,69  €   NPV  scenario  year  4:     115.562,16  €  NPV  scenario  year  5:     132.762,95  €      

Table 6:Comparison NPV’s ‘best implementationyear’ of single action

 

 

Graph 1: Comparison NPV’s ‘best implementationyear’ of single action according to the deterministic method

We  clearly  see  that  the  single  action  can  best  be  implemented  in  the  first  year.  We  can  even  seen  that  if  the  action  is  taken  in  year  3  or  later,  it  does  not  make  sense,  as  the  NPV  is  larger  after  that  year,  rather  

than  smaller.    

In  other  words,  if  the  practical  implementation  of  this  single  action  will  not  be  possible  before  year  3,  due  to  budgeting  problems  for  example,  there  is  no  added  value  to  reserving  the  budget  or  putting  

energy  into  preparing  the  action.  In  this  case  we  can  attempt  to  find  an  alternative,  or  accept  the  current  risk  cost.  

€  0.00    

€  20,000.00    

€  40,000.00    

€  60,000.00    

€  80,000.00    

€  100,000.00    

€  120,000.00    

€  140,000.00    

Year  1:     Year  2:     Year  3:     Year  4:     Year  5:    

Recommended  year  of  implementation  of  action  

Page 11: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 11 of 16

If  the  action  is  simply  changing  the  maintenance  plan  of  the  asset,  then  the  direct  cost  before  and  after  the  action  will  be  different.  Even  if  this  is  the  case  it  is  possible  that  the  choice  is  made  for  an  additional  

single  action.  In  the  function  of  the  calculated  NPV  it  can  be  decided  that  the  altered  maintenance  plan  alone  will  be  sufficient.    

4.2.2. Determination  of  the  failure  pattern  according  to  the  stochastic  method.    

When  using  the  deterministic  method  for  our  calculation,  we  have  used  the  premise  that  we  know  exactly  when  a  failure  will  take  place,  using  the  average  value  MRRF.  We  are  aware  that  in  real  life  this  

will  not  be  possible.    

By  viewing  the  failure  behaviour  of  the  asset  stochastically,  we  attempt  to  incorporate  a  certain  amount  of  accuracy.    

The  failure  behaviour  of  an  asset  can  be  viewed  stochastically  in  several  ways.  Here  we  use  the  most  

common  probability  (distribution)  functions.  In  the  world  of  asset  management  we  usually  work  with  the  following  distribution  functions:    

• Exponential  distribution  

• Standard  distribution  (Bell  curve)  • Weibull  distribution  

The  way  that  the  failure  behaviour  of  the  asset  is  distributed  and  which  probability  function  we  thus  

may  use  depends  on  several  factors.    

Every  probability  function  has  its  specific  shape  and  parameters  that  are  necessary  to  perform  the  

essential  calculations.    

  Exponential  distribution   Standard  distribution   2  parameter  Weibull  distribution  

Curve  shape  

     

Parameters   λ  =  constant  failure  rate  µ  =  average  

σ  =  standard  deviation  β  =  shape  factor  η  =  scale  factor  

Probability  distribution-­‐  function  f(t)  =  

     

Reliability-­‐  function  R(t)  =      

 

Table 7:Functionshape and parameters of the most commonly used probability distributions in asset management

Page 12: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 12 of 16

The  standard  distribution  is  the  most  commonly  used  probability  distribution  in  nature.  Failure  behaviour  of  assets  that  has  a  standard  distribution  has  increasing  failure  intensity.  The  degradation  

symptoms,  amongst  other,  can  thus  be  quickly  described  according  to  a  standard  distribution  function.    

The  Weibull  distribution  can  be  used  as  a  2  parameter  of  3  parameter  distribution.  The  2  parameter  distribution  is  shown  in  table  7.  When  using  a  3  parameter  distribution,  the  third  parameter  is  ‘minimal  

lifetime’,  the  lifetime  in  which  a  component  will  not  fail.    

The  Weibull  distribution  provides  a  good  demonstration  of  situation  in  which  several  failure  mechanisms,  individually,  can  be  the  cause  of  failure  of  a  component  of  an  asset.    

In  the  function  of  the  value  of  the  shape  factor  β,  the  Weibull  distribution  contains  several  other  

probability  distribution  functions,  such  as:  β  =  1:  Exponential  distribution  

β  =  2:  Raleigh  distribution  β  =  3,5:  Normal  distribution  

The  exponential  distribution  is  the  most  simple  to  use  and  to  perform  the  calculations.  This  is  why  this  

function  is  often  used  to  simulate  failure  behaviour.  The  exponential  distribution  is  based  on  ‘random’  failure  behaviour.  Due  to  the  mathematical  simplicity  of  the  exponential  distribution,  we  will  use  the  exponential  distribution  in  this  document.  Please  note  that  the  working  method  for  other  probability  

functions  identical.    

Based  on  the  specific  parameters  of  the  probability  distributions  numerous  failure  patterns  can  be  set  up.    

Similar  to  the  approach  in  the  deterministic  method,  we  will  need  the  number  of  failures  per  years  to  

set  up  the  failure  pattern.    

The  probability  that  the  component  will  still  fulfil  its  function  at  a  certain  time  is  the  reliability  R(t)  of  

this  component  on  that  specific  time.  The  probability  for  the  exponential  distribution  in  the  function  of  the  constant  failure  rate  λ  is  equal  to:  

      (5)  

Where:  t  =  a  chosen  time  expressed  in  X  year  

λ  =  the  (constant)  failure  rate,  equal  to           (6)  

To  map  a  certain  failure  pattern,  we  will  attempt  to  determine  certain  chosen  failure  times,  where  it  is  

of  essence  that  these  failure  patterns  still  meet  the  exponential  distribution  with  a  predetermined  identical  constant  failure  rate  λ.  

This  can  be  done  by  choosing  random  values  of  reliability  R(t)  and  calculating  the  failure  time  several  

times:    

Page 13: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 13 of 16

 

  (7)  

In  this  equation  t  is  time,  expressed  in  year,  where  a  failure  occurs.  In  table  8,  shown  on  the  next  page,  7  of  these  random  times  of  failure  are  illustrated.    

Random  Fail  generator  

1   2   3   4   5   6   7  

Fail  after  x  years   0,37196   1,12248   2,52626   0,06852   2,91534   5,07631   1,04147  

Table 8: Random times of failure, exponentially distributed with MTTF=1,2 (λ = 0,833)

If  these  failures  are  shown  cumulatively,  we  can  see  a  very  specific  failure  pattern,  where  we  can  notice  how  this  failure  pattern  changed  over  time.  The  table  below  shows  the  amount  of  time,  in  years,  has  passed  after  failure  number  7.    

Random  Fail  generator  

1   2   3   4   5   6   7  

Failure  pattern   0,37196   1,49444   4,02070   4,08922   7,00456   12,08087   13,12234  

Table 9: Cumulative failure pattern over time, exponentially distributedwitha MTTF =1,2 (λ = 0,833)

We  can  see  in  this  table  that  the  first  failure  occurs  after  0,37  years.  The  second  failure  occurs  1,49  years  from  now,  etc.  At  the  occurrence  of  failure  number  7,  more  than  13  years  have  passed.    

On  the  basis  of  this  information  we  can  determine  the  number  of  failures  that  will  yearly  take  place,  

according  to  the  specific  failure  pattern.    

Failure  pattern  on  a  yearly  basis  

Year  1   Year  2   Year  3   Year  4   Year  5   Year  6   Year  7  

#  failures  per  year   1   1   0   0   2   0   0  

Table 10: Possible failure pattern, exponentially distributed with MTTF =1,2 (λ = 0,833)

Using  the  data  of  the  cumulative  failure  pattern,  the  table  shown  above  could  be  expanded  with  an  

additional  failure  in  year  13  and  14  (number  6  and  7).    

If  we  now  choose  alternative  random  values  for  R(t),  we  will  find  another  failure  pattern.  It  is  important  to  know  that  all  these  failure  patterns  are  still  valid  for  the  exponential  distribution  according  to  the  

chosen  value  of  λ.  

The  failure  pattern  in  the  table  has  been  set  up  with  an  MTTF  of  1,2  year  (λ  =  1/1,2  =  0,83333).  What  immediately  attracts  our  attention  is  that  the  same  MTTF  of  1,2  years,  a  clearly  different  failure  pattern  

is  found  using  this  method  compared  to  the  deterministic  method.  This  is  logical,  as  there  are  an  infinite  number  of  failure  patterns  possible  with  an  MTTF  of  1,2  years,  and  we  have  shown  2  random  patterns.    

Page 14: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 14 of 16

We  now  have  sufficient  data  as  input  for  the  table  with  the  various  yearly  costs.  The  procedure  of  filling  this  table  is  identical  to  the  procedure  used  in  the  deterministic  method.    

 Total  costs  per  year   Year  1   Year  2   Year  3   Year  4   Year  5  Direct  labour   6.200  €   6.200  €   9.400  €   6.200  €   6.200  €  Direct  material   1.300  €   1.300  €   2.900  €   1.300  €   1.300  €  Total  maintenance  costs   7.500  €   7.500  €   12.300  €   7.500  €   7.500  €  Deprivation  costs  @  MTTF  =  1,2  

25.000  €   25.000  €   0  €   0  €   50.000  €  

Total  cost   32.500  €   32.500  €   12.300  €   7.500  €   57.500  €  PV  cost  @  WACC  7%   30.373,83  €   28.386,76  €   10.040,46  €   5.721,71  €   40.996,71  €  NPV   115.519,47  €          

Table 11: Total cost per year using the stochastic determined failure pattern and MTTF of 1,2 years

If  we  would  generate  an  alternative  failure  pattern,  the  total  NPV,  as  well  as  the  decision  whether  action  would  be  taken,  would  be  different.    

The  total  NPV  and  the  decision  to  implement  an  action,  is  thus  dependent  on  the  randomly  generated  

failure  pattern.  This  can  be  an  aggravating  situation  as  a  basis  for  a  decision.  If  the  NPV’s  that  we  have  calculated  for  the  situation  before  and  after  implementing  an  action  are  very  close,  one  failure  pattern  could  show  a  decision  should  be  implemented,  where  another  failure  pattern  could  show  that  it  should  

not.    

This  is  the  uncertainty  that  rears  its  head  again,  due  to  the  fact  that  we  cannot  predict  the  exact  time  of  the  failure  of  an  asset.    

To  lower  this  uncertainty  we  will  generate  several  different  failure  patterns  by  using  a  Monte  Carlo  

Simulation.  For  each  failure  pattern  we  will  decide  whether  the  action  should  be  implemented  or  not.    

Through  this  method  it  could  be  the  case  that  200  different  failure  patterns  within  an  exponential  

distribution  with  the  same  MTTF,  the  results  show  that  146  times  a  decision  should  be  made  to  implement  the  action,  and  54  times  a  decision  should  be  made  to  not  implement  the  action.  In  other  words,  when  we  take  into  account  the  uncertainty  of  the  exact  failure  pattern,  we  have  a  73&  chance  

that  the  action  leads  to  positive  results  and  a  27%  chance  that  it  does  not.    

 

0.00%  

100.00%  

OK   Not  ok  

Decision  Action  /  No  Action  

Page 15: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 15 of 16

Graph 2:The probability of a positive or negative effect on the decision to implement an action

On  the  basis  of  these  numbers  a  decision  can  be  taken.  Every  organisation  will  have  their  own  decision  

restrictions.  One  organisation  may  implement  a  decision  only  if  the  probability  lies  above  80%,  where  another  organisation  may  decide  to  take  action  if  they  are  55%  certain.    

When  the  NPV  values  are  divergent,  the  percentages  will  also  lie  further  apart,  which  makes  taking  a  

decision  easier.  An  example  would  be  99%  positive  and  1&  negative.    

The  greater  the  number  of  failure  patterns  that  are  simulated,  the  more  accurately  the  percentages  will  approach  reality.    

To  determine  in  what  year  a  single  action  can  best  be  implemented,  a  similar  approach  to  the  

deterministic  method  is  used.  Once  again  the  function  of  the  generated  failure  pattern  will  show  the  ‘best  year  for  implementation’.  By  using  the  Monte  Carlo  Simulation  for  every  failure  pattern,  and  

registering  in  what  year  the  single  action  can  best  be  implemented,  one  can  see  how  many  times  a  specific  year  is  uncovered  as  the  best  implementation  year.  Using  this  data  one  can  then  make  the  decision  when  the  action  should  be  implemented.  In  the  example  shown  below  it  is  recommended  to  

take  the  single  action  in  year  one,  as  this  year  scored  best  in  86  of  the  200  simulations.    

 

Graph 3:Comparison NPV’s ‘best implementation year’of single action according to the stochastic method

5. Decision

Risk  management  does  not  only  imply  the  negation  of  risks,  but  it  can  also  mean  that  some  risks  need  

to  be  accepted.    

This  is  what  can  be  done  using  the  methodology  shown  in  this  paper.  The  method  can  support  and  verify  whether  an  action  should  be  implemented  or  not.  In  other  words,  we  attempt  to  decrease  the  

risk  as  much  as  possible,  or  we  accept  the  risk  as  the  costs  or  action  does  not  outweigh  the  added  value  of  the  action.  The  method  can  also  be  used  to  compare  different  potential  actions  with  reference  to  one  another.    

0  20  40  60  80  100  

Year  1   Year  2   Year  3   Year  4   Year  5  

Recommeded  year  of  action  implementation  

Page 16: Impala competence center   asset management - cost driven risk management

White Paper Cost Driven Risk Management

Page 16 of 16

Every  organization  attempts  to  work  in  a  cost  efficient  way.  The  methods  that  have  been  described  take  into  account  the  cost  aspects  in  decisions.  The  taking  of  actions  to  lower  risks  is  expressed  in  a  ‘financial  

value’,  which  supports  and  simplifies  the  budget  justification  of  asset  managers.    

Risk  analysis  and  cost  driven  decision  making  will  become  increasingly  important  for  asset  managers  to  optimise  the  improvement  plans  and  the  performance  of  their  assets.    

In  closing  a  last  remark  is  made  to  emphasise  the  importance  and  the  use  of  the  methods.  The  methods  

should  be  seen  as  providing  guidance.  It  will  never  be  possible  to  determine  the  exact  failure  pattern  of  an  asset,  and  we  can  thus  never  take  a  decision  that  we  are  absolutely  certain  about.    

The  use  of  the  stochastic  method  to  determine  failure  behaviour,  and  the  use  of  Monte  Carlo  

Simulation  to  generate  several  failure  pattern,  can  give  us  more  substantive  guidance  than  the  use  of  the  deterministic  method  which  is  solely  based  on  the  failure  pattern  according  the  MTTF.